language models Reliable LLM reasoning Telling correct reasoning from errors using how a model's internal representations change across layers. Partial-LoRA Fine-tuning with up to 87% fewer trainable parameters while maintaining or improving accuracy. Model editing with task vectors Separating shared and task-specific parts of task vectors for more targeted model editing. applied AI Multi-agent LLM deliberation for AI policy An LLM multi-agent framework with RAG context and generated personas, used to simulate stakeholder deliberation. Electricity-demand forecasting Probabilistic forecasting of local demand and the state's annual peaks, informing battery and solar use. Corneal-disease detection Leading a healthcare vision project from clinical data collection to weakly supervised detection. robotics and perception Robot localisation and sensor fusion Uncertainty-aware odometry and visual, inertial and kinematic fusion for robots.
